Sampling Process:

The message signal is usually analog

in nature, as in a speech signal or
video signal
It has to be converted into digital
form before it can be transmitted by
digital means.
Sampling Process:
The sampling processing is the first
process preformed in analog-to-digital
conversion.
In the sampling process, a continuous-time
signal is converted into a discrete-time
signal by measuring the signal at periodic
instants of time.

Sampling Process:
For the sampling process to be of
practical utility, it is necessary that we
choose the sampling rate properly
So the discrete-time resulting from
the process uniquely defines the
original continuous-time signal.
Sampling Theorem:
Let the signal
) (t x
be band limited with
bandwidth
W
i.e., let
0 ) ( f X . W f >
for
Let
) (t x be sampled at multiples of some
basic sampling interval
S
T , where
W
T
S
2
1
s
to yield the sequence
( ) | |

= n
S
nT x
Then it is
possible to reconstruct the original
Sampling Theorem:
signal ) (t x from the sampled values by the
reconstruction formula:
( ) ( )] 2 [ sin 2 ) (
S
n
S S
nT t W c nT x T W t x
' '
=

=
W
'
Where ( ) is any arbitrary number that
that satisfies .
1
W
T
W W
S
s
'
s
Sampling Theorem:
special case where
W
T
S
2
1
=
the
reconstruction relation simplifies to:


=

=
(

|
.
|

\
|

|
.
|

\
|
=
|
|
.
|

\
|
=
n
S
n
S
W
n
t W c
W
n
x n
T
t
c nT x t x
2
2 sin
2
sin ) ( ) (
Let
) (t x
o
denote the result of the sampling
original signal by impulses at
S
nT time instants.
Sampling Theorem:
Then:

=
=
n
S S
nT t nT x t x ) ( ) ( ) ( o
o
We can write
) (t x
o
as:

=
=
n
S
nT t t x t x ) ( ) ( ) ( o
o

=
=
n
S S
nT t nT x t x ) ( ) ( ) ( o
o

=
=
n
S
nT t t x t x ) ( ) ( ) ( o
o
Sampling Theorem:
f
W -W -fc fc 0
. . . .
. . . .
W f
c

o c
x f
) ( f x
o
f
-W W
0
o
x
) ( f x
Figure (1):
Signal spectra for low pass sampling.
(a) Assumed spectrum for x(t).
(b) Spectrum of sampled signal.
Sampling Theorem:
Now if we find the Fourier transform of
both sides of the above relation and apply
the dual of the convolution theorem to the
right-hand side, we obtain:
( ) ( ) ( ) .....(4)
S
n
X f X f F t nT
o
o

=
(
= -
(

Sampling Theorem:
By using Fourier Transform we obtain:
(

= n
S
nT t F ) ( o
1
....(5)
n
S S
n
f
T T
o

=
| |
=
|
\ .

By substituting equation (5) into equation (4),

we obtain:

=
|
|
.
|

\
|
- =
n
S S
T
n
f
T
f X f X o
o
1
) ( ) (

=
|
|
.
|

\
|
=
n
S S
T
n
f X
T
1
Sampling Theorem:
Where in the last step we have employed the
convolution property of the impulse signal.
This relation shows that ) ( f X
o
, the Fourier
transform of the impulse-sampled signal is a
replication of the Fourier transform of the
original signal at a
S
T
1
rate.
Figure (1) shows this situation.
Sampling Theorem:
Now if
W
T
S
2
1
> then the replicated spectrum of
) (t x
overlaps, and reconstruction of the original
signal is not possible. This type of distortion
that results from under-sampling is known as
aliasing error or aliasing distortion.
Sampling Theorem:
However, if
W
T
S
2
1
s no overlap occurs, and by
employing an appropriate filter we can
reconstruct the original signal back. To obtain
the original signal back, it is sufficient to filter
the sampled signal by a low pass filter with
frequency response characteristic
Sampling Theorem:
S
T f H = ) (
W f <
0 ) ( = f H
W
T
f
S
>
1
1.
for
.
2.
for
For
W
T
f W
S
< s
1
, the filter can have any
characteristics that make its implementation easy.
Of course, one obvious (though not practical)
choice is an ideal low pass filter with bandwidth
W
'
W
' W
T
W W
S
<
'
s
1
where
satisfies , i.e.
Sampling Theorem:
|
.
|

\
|
'
H =
W
f
T f H
S
2
) (
With this choice we have:
|
.
|

\
|
'
H =
W
f
T f X f X
S
2
) ( ) (
o
Taking inverse Fourier transform of both sides,
we obtain:
( ) t W c T W t x t x
S
' '
- = 2 sin 2 ) ( ) (
o
Sampling Theorem:
( ) ( ) ( ) t W c T W nT t nT x
S
n
S S
' '
- |
.
|

\
|
=

=
2 sin 2 o
( ) ( ) | |

=

' '
=
n
S S S
nT t W c nT x T W 2 sin 2
This relation shows that if we use sine functions
for interpolation of the sampled values, we can
reconstruct the original signal perfectly.
Sampling Theorem:
The sampling rate
W
f
S
2
1
=
is the minimum
sampling rate at which no aliasing occurs.
This sampling rate is known as the Nyquist
sampling rate.
If sampling is done at the Nyquist rate,
then the only choice for the reconstruction
filter is an ideal low pass filter and .
2
1
S
T
W W = =
'
Sampling Theorem:
Then:
( )

=

|
.
|

\
|
=
n
n Wt c
W
n
x t x 2 sin
2
) (
( )

=
|
|
.
|

\
|
=
n
S
S
n
T
t
c nT x t x sin ) (
In practical systems, sampling is done at a rate
higher than the Nyquist rate. This allows for
the reconstruction filter to be realizable and
easier to build.
Sampling Theorem:
In such cases the distance between two adjacent
replicated spectra in the frequency domain; i.e.
W f W W
T
S
S
2
1
=
|
|
.
|

\
|

, is known as the guard band.

Note that there exists a strong similarity
between our development of the sampling
theorem and our previous development of the
Fourier transform for periodic signals
(or Fourier series).
Sampling Theorem:
In the Fourier transform for periodic signals,
we started with a time periodic signal and
showed that its Fourier transform consists of
a sequence of impulses.
Therefore, to define the signal, it was enough
to give the weights of these Impulses
(Fourier series coefficients).
Sampling Theorem:
In the sampling theorem, we started with an
impulse-sampled signal, or a sequence of
impulses in the time domain, and showed that
the Fourier transform is a periodic function in
the frequency domain. Here again, the values
of the samples are enough to define the signal
completely.
Sampling Theorem:
This similarity is a consequence of the duality
between the time and frequency domains and
the fact that both the Fourier series expansion
and reconstruction from samples are orthogonal
expansions, one in terms of the exponential
signals and the other in terms of the sine
Functions.
Analog Pulse Modulation:
In the sampling theory section we show that
continuous band limited signals can be
represented by a sequence of discrete samples
and that the continuous signal can be
reconstructed with negligible error if the
sampling rate is sufficiently high.
Consideration of the sampled signals leads us
to the topic of the pulse modulation.
Analog Pulse Modulation:
Pulse modulation can be either analog, in which
some attribute of a pulse varies continuously in
one-to-one correspondence with sample value,
or digital, in which some attribute of a pulse can
take on a certain value from a set of allowable
values.
Analog Pulse Modulation:
t
t
t
t
Analog
Signal
(Samples)
PAM Signal
PWM
Signal
PPM
Signal
Ts 2Ts
9Ts 0
Figure (2): illustration of
PAM, PWM, and PPM
Analog Pulse Modulation:
As mentioned Analog Pulse Modulation
results when some attribute of a pulse values
continuously in one-to-one correspondence
with a sample value. There are three pulse
attributes that can be readily varied:
Amplitude, Width, and Position.
Analog Pulse Modulation:
These lead to pulse amplitude modulation
(PAM), pulse width modulation (PWM), and
pulse position modulation (PPM), as illustrated
in figure (2).
Pulse Amplitude Modulation:
A (PAM) waveform consists of a sequence of
a flat-topped pulses designating sample value.
The amplitude of each pulse corresponds to the
value of the message signal at the leading edge
of the pulse.
The essential difference between PAM and
sampling operation is that in PAM we allow the
sampling pulse to have finite width.
Pulse Amplitude Modulation:
The finite-width pulse can be generated from
impulse-train sampling function by passing
the impulse-train sample through a holding
network as shown in figure (3). The holding
network transforms the impulse function
samples, given by:
( ) ( )

=
=
n
s S
nT t nT x t x o
o
) (
Pulse Amplitude Modulation:
(a)
Input
PAM Output
) (t h
(b)
) (t h
0
t
t
Slope= -
(d)
f
) ( f H Z
t
t
2/ 1/
-1/ -2/
f
0
(c)
) ( f H
2/ 1/
-1/ -2/
f
0
(c)
) ( f H
Figure (3): Generation of PAM.
(a) Holding network.
(b) Impulse response of holding network.
(c) Amplitude response of holding network.
(d) Phase response of holding network.
Pulse Amplitude Modulation:
From figure (2) a PAM signal can be written as:
( )

=
(
(
(
(

|
.
|

\
|
+
[ =
n
S
S PAM
nT t
nT x t x
t
t
2
1
) (
The waveform is generated by placing the
impulse function in (11) on the output of a
holding network having the impulse response.
Pulse Amplitude Modulation:
(
(
(
(

|
.
|

\
|
+
[ =
t
t
2
1
) (
S
nT t
t h
And the transfer function is:
( )
t t
t t
f j
e f c f H

= sin ) (
Pulse Amplitude Modulation:
Since the holding network dose not have a
constant amplitude response over the bandwidth
of
) (t x
t
, unless of course the pulse width

is sufficiently narrow, amplitude distortion
results. This amplitude distortion can be
removed by passing the samples, prior to
reconstruction of
) (t x
Pulse Amplitude Modulation:
through a filter having an amplitude response
equal to

) ( 1 f H
, over the bandwidth of ). (t x
.
Since the phase response of the holding
network is linear, the effect is a time delay and
can usually be neglected.
Pulse Width Modulation:
A (PWM) waveform, as illustrated in figure (2),
consists the sequence of pulse width each pulse
having a width proportional to the values of the
a message signal at the sampling instants.
If the message is (0) at the sampling time, the
width of the (PWM) pulse is
.
2
1
S
T
Pulse Width Modulation:
Thus, pulse widths less than
S
T
2
1
correspond
to negative sample values and the pulse widths
greater than correspond to positive sample
S
T
2
1
values.
PWM is seldom used in modern communications
systems.
Pulse Width Modulation:
PWM is used extensively for DC motor
control in which motor speed is proportional
to the width of the pulses.
Since thee pulses have equal amplitude, the
energy in a given pulse is proportional to the
pulse width. Thus, the sample values can be
recovered from a PWM waveform by low pass
filtering.
Pulse Position Modulation:
A (PPM) signal consists of a sequence of
pulses in which the pulse displacement from
a specified time reference is proportional to
the sample values of the information-bearing
signal.
A (PPM) signal is illustrated in figure (2),
and can be represented by the expression:
Pulse Position Modulation:

=
=
n
n
t t g t x ) ( ) (
Where
) (t g
represents the shape of the
individual pulses, and occurrence times
n
t
are related to the values of the message signal
) (t x
S
nT
at the sampling instants
, as discussed
previously.
Pulse Position Modulation:
The spectrum of a PPM signal is very similar to
the spectrum of a PWM signal.
If the time axis is slotted so that a given range
of sample values is associated with each slot,
the pulse positions are quantized and pulse is
assigned to given slot depending on the sample
value.
Pulse Position Modulation:
Slots are non-overlapping and are therefore
orthogonal.
If a given sample value is assigned to one
of (M) slot, the result is (M-ary) orthogonal
communications. PPM is finding new
applications in area of ultra-wideband
communications.